Structural mechanisms for ATP-mediated inhibition of human NLRP6 inflammasome Authors: Zhen Cui, Qingyu Sheng, Minsoo Son, Young Ah Goo, Chen Shen Published: 2026 ABSTRACT Inflammasomes ignite innate immune defense in response to infectious pathogens and noninfectious dangers, primarily through sensors composed of nucleotide-binding domain (NBD), leucine-rich repeat (LRR)-containing (NLR) family proteins. NLRP6 is an inflammasome sensor that plays critical roles in regulating intestinal inflammation, and its overactivation is linked to autoinflammatory diseases such as inflammatory bowel disease. However, how NLRP6 is maintained in an inhibited structure is unknown. Here we report two cryogenic-electron microscopy structures of human NLRP6 monomer in the adenosine-5′-triphosphate (ATP)/NBD-bound (twisted conformation) and unbound (extended conformation) states. The ATP-binding event connects and compacts the NACHT subdomains and the LRR domain, thus maintaining NLRP6 in an inhibitory conformation. NBD interacts directly with helical domain 1, winged-helix domain and helical domain 2, further contributing to the autoinhibition. Disruption of ATP binding and NBD interactions unleashes the NLRP6 inflammasome activation in the cellular study. The structural comparison between twisted and extended conformations reveals that the rearrangement of an NLRP6-specific acidic loop modulates NLRP6 activity. Although ATP-binding of NLRP6 and MCC950 (a potent NLRP3 inhibitor)-binding of NLRP3 share a similar interaction location in the structures, MCC950 does not inhibit NLRP6 in cells. Together, our data reveal the ATP-mediated cooperative inhibition mechanism of NLRP6 and provide insight into the therapeutic intervention of NLRP6-related autoinflammatory disorders. Main The nucleotide-binding domain (NBD) and leucine-rich repeat (LRR)-containing proteins (NLRs) represent one main type of cytosolic pattern recognition receptors that detect exogenous pathogen invasion, endogenous damage-associated stimuli or homeostasis-altering molecular patterns1–4. Once activated, NLR proteins primarily form multiprotein complexes in the cytoplasm named inflammasomes5–9. Assembly of the inflammasome machinery leads to the activation of inflammatory caspases such as caspase-1, which processes type I interleukins (ILs) such as IL-1β or IL-18 and gasdermin D (GSDMD), resulting in the release of these cytokines and pyroptotic cell death5,10–12. NLRP6 is a tissue-specific NLR protein expressed predominantly in the intestine and liver13,14 that plays multifaceted roles in innate immune defense and intestinal homeostasis15–20. The inflammasome function of NLRP6 was reported to defend against mouse hepatitis virus infection in the liver21, exacerbate the symptoms of Listeria monocytogenes infection in mice due to excessive inflammation22, sense the metabolites from commensal microbiota for the production of IL-18 and anti-microbial peptides23, and regulate goblet cell mucus secretion24,25. Most recently, the NLRP6 inflammasome was also implicated in playing crucial roles in sensing endolysosomal damage in intestinal epithelial cells26 and detecting intestinal protozoa infection through monitoring microbial sphingolipids27. NLRP6 has an N-terminal pyrin domain (PYD) that is responsible for signal transduction through the recruitment of downstream adapter protein ASC; a central NACHT domain that harbors NBD, helical domain 1 (HD1), winged-helix domain (WHD) and HD2, and a C-terminal scaffold LRR domain (Fig. 1a). Our previous studies suggested that both NLRP6 PYD and full-length protein nucleate the polymerization of ASC through homotypic PYD–PYD interaction, confirming the assembly of NLRP6 inflammasome in vitro28, which is further supported by recent cellular studies showing that NLRP6 is one of the dominant NLRP proteins that induce robust inflammasome signaling in reconstituted cell lines29,30. Meanwhile, double-stranded RNA (dsRNA), the molecular pattern derived from RNA viruses, and lipoteichoic acid, the molecular pattern derived from gram-positive bacteria, have been reported as bona fide ligands to trigger NLRP6 inflammasome signaling through the formation of biomolecular condensates—a unique mode of higher-order assembly21,22. Although genetic and biochemical studies have endorsed the important function of NLRP6 in the innate immune system, the structural mechanisms of NLRP6 inhibition are unknown, which hinders therapeutic strategies to target this important inflammasome sensor.Fig. 1Cryo-EM structures of human NLRP6 monomer.a, Domain architecture of human NLRP6. Domains/subdomains are color coded, and domain boundaries are labeled. b, Representative 2D class averages from cryo-EM data processing showing two distinct conformations. Scale bars, 90 Å. c, Structure of human NLRP6 in the ATP/NBD-bound state (twisted conformation, related to the state 1 in b) and the corresponding cryo-EM map. Spheres, ATP. d, Structure of human NLRP6 in the ATP/NBD-unbound state (extended conformation, related to the state 2 in b) and the corresponding cryo-EM map. In c and d, different domains and structural elements are color coded. Dashed boxes: domains missing in the solved structures. The acidic loop is colored in gray, and the location is indicated in the domain arrangements. Cryo-EM structures of human NLRP6 monomer. a, Domain architecture of human NLRP6. Domains/subdomains are color coded, and domain boundaries are labeled. b, Representative 2D class averages from cryo-EM data processing showing two distinct conformations. Scale bars, 90 Å. c, Structure of human NLRP6 in the ATP/NBD-bound state (twisted conformation, related to the state 1 in b) and the corresponding cryo-EM map. Spheres, ATP. d, Structure of human NLRP6 in the ATP/NBD-unbound state (extended conformation, related to the state 2 in b) and the corresponding cryo-EM map. In c and d, different domains and structural elements are color coded. Dashed boxes: domains missing in the solved structures. The acidic loop is colored in gray, and the location is indicated in the domain arrangements. In this study, we determine two cryogenic electron microscopy (EM) (cryo-EM) structures of human NLRP6 monomer in the ATP/NBD-bound (twisted conformation) and ATP/NBD-unbound states (extended conformation). We made the unprecedented discovery that ATP restricts NLRP6 inflammasome activity by binding to a noncanonical pocket in the twisted conformation, which we further validated in biochemical and cellular studies. In addition, the presence and absence of NBD, and structural rearrangements of a unique acidic loop between two different conformations critically modulate NLRP6 inflammasome activity. Our work uncovered the molecular mechanisms of NLRP6 inhibition and regulation mediated by ATP and different structural elements. Results Structure determination of human NLRP6 under two conformational states Previous studies reported that recombinant human NLRP6 protein tends to form higher-order oligomers or biomolecular condensates21,31. To tackle the aggregation-prone behavior of NLRP6, we fused an N-terminal His6 (6×Histidine)-maltose binding protein (MBP) tag to increase the protein solubility along with a noncleavable linker between MBP and NLRP6 (refs. 28,32). We expressed the final construct pFastBac-MBP-NLRP6 (14–892) utilizing the Bac-to-Bac expression system by infecting Spodoptera frugiperda (Sf)9 cells with baculovirus and purified the target protein using Ni-nitrilotriacetic acid (NTA) agarose. The gel-filtration profile exhibits a void peak, with particles shown as large assemblies under negative stain EM, and a 12.2-ml peak having smaller homogeneous particles (Extended Data Fig. 1a–d). To increase the purity of the protein sample for further structural study, we re-ran the size-exclusion chromatography (SEC) of the sample fractionated from peak 1 in the original chromatography (Extended Data Fig. 1e). The peak fractions were collected and concentrated for the subsequent cryo-EM data acquisition (Extended Data Fig. 1f). A cryo-EM dataset of 14,261 movies was collected on a Titan Krios equipped with a Falcon4EC direct detector. Surprisingly, the two-dimensional (2D) classification showed two main classes with distinct secondary structural features during the data processing using cryoSPARC33 (Fig. 1b). After three-dimensional reconstruction (Extended Data Fig. 2a), we successfully resolved the cryo-EM maps of two different states of human NLRP6 monomer: state 1 with a twisted conformation at 3.7 Å (Fig. 1c, Table 1 and Extended Data Fig. 2b–d) and state 2 with an extended conformation at 3.8 Å resolution (Fig. 1d, Table 1 and Extended Data Fig. 2e–g). For model building of the twisted conformation, we found that the AlphaFold34 predicted HD1, WHD, HD2 and LRR—typical domains in NLR superfamily proteins—fit very well into the EM density. The NBD model was built and modified separately due to a unique conformation, based on the individual NBD domain in the AlphaFold predicted structure. The MBP tag and PYD domain were missing (Fig. 1c and Extended Data Fig. 3a,b). More strikingly, we observed clear density located at a cleft that tightly connects different NLRP6 domains NBD, HD1, HD2 and LRR, which can be readily assigned as an ATP molecule (Fig. 1c and Extended Data Fig. 3c,d). To further confirm whether insect cell-purified NLRP6 binds to endogenous ATP, the protein sample (at 0.5 mg ml−1) was analyzed following trichloroacetic acid (TCA) quenching, mixed-mode anion exchange solid phase extraction (MAX–SPE) enrichment, and hydrophilic interaction liquid chromatography–Orbitrap high-resolution liquid chromatography–mass spectrometry (LC–MS). The ATP signal detected in the protein sample showed matched chromatographic retention time with the ATP standard under the same analytical conditions (Extended Data Fig. 3e). In addition, the precursor ion in the protein-derived extract displayed an MS1 isotopic pattern consistent with that of ATP and matched the expected ATP ion cluster (Extended Data Fig. 3f). Further confirmation was obtained by parallel reaction monitoring analysis, in which the protein-derived ATP signal exhibited a fragment ion pattern highly similar to that of the ATP standard, with four matched product ions and a cosine similarity score of 0.9817 (Extended Data Fig. 3g). Hence, the agreement in chromatographic elution, precursor isotopic distribution and tandem mass spectral fragmentation supports the presence of ATP in the protein-derived sample and indicates that ATP can be recovered from the purified NLRP6 following extraction. For model building of the extended conformation, HD1/WHD/HD2/LRR domains were also well resolved, whereas both PYD and NBD are absent in the cryo-EM structure (Fig. 1d and Extended Data Fig. 4). By analyzing the AlphaFold predicted structure (Extended Data Fig. 5a–c) and the sequence alignment among different NLRP proteins (Extended Data Fig. 5d), we located a unique loop region (residues 586–616) with abundant glutamates in hNLRP6 that are not seen in other NLRs, which is designated as the acidic loop. This unique acidic loop is seen in both twisted and extended NLRP6 structures, albeit with different lengths (Fig. 1c,d). Different from the AlphaFold prediction showing the loop is highly flexible (Extended Data Fig. 5a–c), in the extended conformation, the acidic loop establishes extensive interaction with the LRR domain (Fig. 1d). However, in the twisted conformation, the acidic loop is much shorter, and alternatively coordinates the interaction between ATP and other domains, such as HD2 and LRR (Extended Data Fig. 3d). Moreover, the presence of NBD and the ‘molecular glue’ ATP in the twisted state causes NLRP6 to adopt a more compact conformation compared to the extended state. Taken together, our structural observations raise interesting questions as to whether these unique molecular features regulate the activity of NLRP6 inflammasome.Table 1Cryo-EM data collection, refinement and validation statisticsHuman NLRP6 extended conformation(EMD 70555)(PDB 9OK7)Human NLRP6 twisted conformation(EMD 70556)(PDB 9OK8)Data collection and processingMagnification165,000Voltage (kV)300Electron exposure (e−/Å2)48.44Defocus range (μm)−0.6 to −2.5Pixel size (Å)0.73Symmetry imposedC1Initial particle images (no.)7,565,037Final particle images (no.)76,219153,064Map resolution (Å)3.793.69Fourier shell correlation threshold0.1430.143RefinementInitial model usedAF-P59044-F1AF-P59044-F1Model resolution (Å)4.24.3 Fourier shell correlation threshold0.50.5Map sharpening B factor (Å2)−165.8−186.8Model composition Nonhydrogen atoms3,8174,636 Protein residues484583 Ligands0ATP:1B factors (Å2) Protein118.0091.45 LigandN/A77.98Root mean square deviations Bond lengths (Å)0.0030.005 Bond angles (°)0.6030.947ValidationMolProbity score2.192.57Clashscore16.2927.96Poor rotamers (%)00Ramachandran plot Favored (%)92.4186.26 Allowed (%)7.5913.74 Disallowed (%)00 Table 1Cryo-EM data collection, refinement and validation statisticsHuman NLRP6 extended conformation(EMD 70555)(PDB 9OK7)Human NLRP6 twisted conformation(EMD 70556)(PDB 9OK8)Data collection and processingMagnification165,000Voltage (kV)300Electron exposure (e−/Å2)48.44Defocus range (μm)−0.6 to −2.5Pixel size (Å)0.73Symmetry imposedC1Initial particle images (no.)7,565,037Final particle images (no.)76,219153,064Map resolution (Å)3.793.69Fourier shell correlation threshold0.1430.143RefinementInitial model usedAF-P59044-F1AF-P59044-F1Model resolution (Å)4.24.3 Fourier shell correlation threshold0.50.5Map sharpening B factor (Å2)−165.8−186.8Model composition Nonhydrogen atoms3,8174,636 Protein residues484583 Ligands0ATP:1B factors (Å2) Protein118.0091.45 LigandN/A77.98Root mean square deviations Bond lengths (Å)0.0030.005 Bond angles (°)0.6030.947ValidationMolProbity score2.192.57Clashscore16.2927.96Poor rotamers (%)00Ramachandran plot Favored (%)92.4186.26 Allowed (%)7.5913.74 Disallowed (%)00 Cryo-EM data collection, refinement and validation statistics ATP restricts the activity of NLRP6 inflammasome ATP hydrolysis is linked tightly to the regulation of NLR inflammasomes35–39. It has been speculated that NLRs can switch between adenosine diphosphate (ADP)-bound inactive state and ATP-bound active state39,40. Indeed, previously published structures of the inactive state of other NLR family proteins showed that the NBD domain is coordinated by ADP (in the case of NLRP3 and NLRC4)39,41–43 or adenosine-5′-[γ-thio]triphosphate (ATP-γS) (in the case of NLRP1)44 to interact with the HD1 domain and WHD domain (Fig. 2a). The binding site of ADP or ATP-γS is designated as the canonical ATPase catalytic site. In our determined NLRP6 twisted conformation, no nucleotide was observed in the potential canonical pocket (Fig. 2b). Instead, we detected that ATP molecule binds to a distinct site and coordinates almost all domains of NLRP6 (Fig. 2b and Extended Data Fig. 3d). Based on this unprecedented discovery of nucleotide binding in NLRP6 monomer, we want to first explore whether NLRP6 still has ATPase activity. Recombinant production of human NLRP6 yielded a void fraction and a low-molecular weight fraction, corresponding to the higher-order assembly and the monomeric state, respectively (Extended Data Fig. 1a–d). We used the malachite green phosphate assay to assess free phosphate production for both forms. The results showed that monomeric NLRP6 has a low basal level of ATP hydrolysis, whereas the ATPase activity of oligomeric NLRP6 is notably higher (Extended Data Fig. 6a,b), which is similar to what has been shown for NLRP3 (ref. 36). This observation is further confirmed in the protein concentration-dependent and time-dependent ATPase assay for the void and monomeric protein samples (Extended Data Fig. 6c,d). These data suggest that NLRP6 is an ATPase and the conversion from monomeric to oligomeric NLRP6 enhances ATP hydrolysis. To study the source of ATPase in NLRP6, we then performed the sequence alignment of Walker A and Walker B motifs, which are highly conserved amino acid sequences within the NACHT domain of NLR proteins and are required for nucleotide binding and ATP hydrolysis36. The results showed NLRP6 shares the consensus sequences GxxxxGK(T/S) (x is any amino acid) in Walker A and hhhhDGxDE (h is a hydrophobic residue) in Walker B with other NLRs (Extended Data Fig. 6e). However, despite the sequence conservation, the mutagenesis on Walker A/B motif residues did not diminish the ATP hydrolysis in NLRP6 monomer (Fig. 2c). Given that we identified an uncharacterized ATP-binding pocket in the twisted conformation, we want to elucidate whether this unique ATP-binding event contributes to ATP hydrolysis. The detailed structural analysis showed that ATP is embedded in a strong binding environment consisting of majorly electrostatic interactions with NLRP6 (Fig. 2d). The adenosine ring establishes hydrogen bonds with the backbone of R650, the side chain of D655 in the LRR domain, E611 in the acidic loop and a hydrophobic interaction with Y623 in the HD2 domain. The ribose ring of ATP coordinates R545 and E620 in the HD2 domain and interacts potentially with E379 in the HD1 domain. The α-phosphate group of ATP generates van der Waals interactions with A383 in the HD1 domain. The γ-phosphate group establishes a strong hydrogen-bonding network with E609 and E613 in the acidic loop, H539 in the HD2 domain, backbone of A330 in the NBD domain and E504 in the WHD domain (Fig. 2d). To evaluate the function of ATP binding, we expressed and purified the structure-guided mutant proteins (Extended Data Fig. 7a,b) and assessed them biochemically using the ATPase assay mentioned above. For the assessment of NLRP6 WT and mutant monomers (Fig. 2e), H539L and Y623A mutations boosted ATP hydrolysis activity substantially compared to the WT protein, with an enhancement of 2.0-fold and 3.8-fold, respectively. E620A, R650G and E379A had a moderate effect in increasing the enzyme activity. E504A and D655A showed comparable activity to that of NLRP6 wild type (WT) (Fig. 2e). Overall, the above enzymatic assays implicated neither the canonical site (Walker motifs, Fig. 2c) nor the newly identified pocket (Fig. 2d,e) executes ATP hydrolysis in monomeric NLRP6. It should be mentioned that, compared to WT protein, E611A and E613A in the acidic loop slightly attenuated the ATPase activity of the monomer (Fig. 2e), but not the oligomeric form (Extended Data Fig. 7c). However, how E611 and E613 contribute to the basal level of ATP hydrolysis in monomer is structurally unknown.Fig. 2Structural insights into ATP-mediated suppression of NLRP6 inflammasome.a, Previously published structures of monomeric mouse NLRC4 (left), human NLRP1 (middle) and human NLRP3 (right) under inhibitory conformations. The NLR structures displayed here were adapted from the original PDBs with the ID codes shown. ADP/ATP-γS-binding at the canonical ATPase pocket is highlighted. b, Resolved structure of NLRP6 twisted conformation. No nucleotide is observed at the potential canonical ATPase pocket. ATP-binding to a new pocket is highlighted. Domains are color coded in a and b. c, ATP hydrolysis activity of the monomeric fractions of NLRP6 WT and Walker A/B motif mutants. d, Detailed structural analysis of the noncanonical ATP-binding pocket in the twisted conformation of NLRP6 monomer. Domains and interface residues are color coded. The ATP molecule is highlighted in stick representation. e, ATP hydrolysis activity of the monomeric fractions of NLRP6 WT and ATP-binding interface mutants. f, Left: anti-Flag immunoblots to detect the expression of inflammasome components Flag-NLRP6, Flag-ASC and Flag-caspase-1. β-actin expression was used as a loading control. Right: anti-caspase-1 immunoblots to assess the activity of NLRP6 inflammasomes by detecting the auto-processing product of p20 fragment. Different proteins and caspase-1 fragments are labeled. Caspase-1 alone and caspase-1 + ASC were used to assess the level of caspase-1 autoactivation. g, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT and ATP-binding mutants inflammasomes in HEK-293T cells. Numbers at the bottom of caspase-1 blot suggest the intensity (I) ratio of p20 bands for several ATP-binding pocket mutants to the WT (Imutant/IWT). The band intensity was measured using ImageJ55. These experiments were repeated three independent times. In c and e, data are presented as mean ± s.e.m. from three technical repeats. The concentration of free phosphate is generated by converting A620 absorbance using the standard curve in Extended Data Fig. 6b. Statistical significance was determined by one-way analysis of variance and Tukey’s honest significant difference post hoc analysis. NS, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.Source data Structural insights into ATP-mediated suppression of NLRP6 inflammasome. a, Previously published structures of monomeric mouse NLRC4 (left), human NLRP1 (middle) and human NLRP3 (right) under inhibitory conformations. The NLR structures displayed here were adapted from the original PDBs with the ID codes shown. ADP/ATP-γS-binding at the canonical ATPase pocket is highlighted. b, Resolved structure of NLRP6 twisted conformation. No nucleotide is observed at the potential canonical ATPase pocket. ATP-binding to a new pocket is highlighted. Domains are color coded in a and b. c, ATP hydrolysis activity of the monomeric fractions of NLRP6 WT and Walker A/B motif mutants. d, Detailed structural analysis of the noncanonical ATP-binding pocket in the twisted conformation of NLRP6 monomer. Domains and interface residues are color coded. The ATP molecule is highlighted in stick representation. e, ATP hydrolysis activity of the monomeric fractions of NLRP6 WT and ATP-binding interface mutants. f, Left: anti-Flag immunoblots to detect the expression of inflammasome components Flag-NLRP6, Flag-ASC and Flag-caspase-1. β-actin expression was used as a loading control. Right: anti-caspase-1 immunoblots to assess the activity of NLRP6 inflammasomes by detecting the auto-processing product of p20 fragment. Different proteins and caspase-1 fragments are labeled. Caspase-1 alone and caspase-1 + ASC were used to assess the level of caspase-1 autoactivation. g, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT and ATP-binding mutants inflammasomes in HEK-293T cells. Numbers at the bottom of caspase-1 blot suggest the intensity (I) ratio of p20 bands for several ATP-binding pocket mutants to the WT (Imutant/IWT). The band intensity was measured using ImageJ55. These experiments were repeated three independent times. In c and e, data are presented as mean ± s.e.m. from three technical repeats. The concentration of free phosphate is generated by converting A620 absorbance using the standard curve in Extended Data Fig. 6b. Statistical significance was determined by one-way analysis of variance and Tukey’s honest significant difference post hoc analysis. NS, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. Source data Besides promoting ATP hydrolysis, our SEC results showed that Y623A, H539L, E620A and R650G exhibited higher void:monomer ratio than WT, derived from the integration of individual peaks (Extended Data Fig. 7d). Because the oligomeric state of NLRs was reported to be the active conformation40,45,46, these biochemical data suggest that disruption of ATP binding in the newly resolved pocket triggers higher-order assembly of NLRP6, which generates the active form of NLRP6 inflammasome that has elevated ATPase activity. Meanwhile, the void fractions of NLRP6 WT and all mutants showed comparable high-level ATPase activity, demonstrating that they are fully activated forms of NLRP6 that are not distinguishable from each other (Extended Data Fig. 7c). To validate that ATP-binding indeed auto-inhibits the NLRP6 inflammasome in a cellular setting, we expressed Flag-tagged NLRP6 WT and the mutant forms, along with ASC and caspase-1 in human embryonic kidney (HEK)-293T cells. Our goal was to determine whether ATP-binding pocket mutations trigger the spontaneous activation of NLRP6 inflammasome by monitoring the level of caspase-1 self-cleavage (p20 fragment)—a hallmark indicator of NLR inflammasome activation5,20,39. Under our experimental conditions, reconstitution of WT NLRP6 inflammasome resulted in a low basal level of p20 fragment production, which indicates that NLRP6 is in an inhibited state even when we ectopically overexpress the inflammasome components (Fig. 2f,g). For the mutants we tested, H539L, E620A, Y623A, R545A in the HD2 domain, E611A and E613A in the acidic loop and A383G in the HD1 domain promoted caspase-1 cleavage substantially, suggesting the effective induction of NLRP6 inflammasome. E504A in the WHD, D655A in the LRR, E379A in the HD1 and A330G in the NBD increased NLRP6 activity slightly, whereas E609A behaved similarly to WT (Fig. 2g). It is worth noting that Y623A enhanced caspase-1 cleavage and promoted ATP hydrolysis, whereas E611A/E613A enhanced caspase-1 cleavage but slightly decreased the ATPase activity in the monomeric fraction (Fig. 2e,g). The void:monomer ratio calculated from the gel-filtration profiles hinted that E611A and E613A mutants do not readily promote oligomerization, having a ratio comparable to that of the WT protein, whereas Y623A is a much stronger booster of the oligomeric form (Extended Data Fig. 7d). These results made us speculate that, when purified in vitro, E611A and E613A by themselves still exhibit dominantly monomers at a potential ‘checkpoint’ state, which has low basal ATPase activity. Although, in the reconstituted cell line, the ‘checkpoint’ state was further stimulated by the cooperative effect of all inflammasome components (NLRP6E611A/NLRP6E613A, ASC and caspase-1 together) and p20 cleavage was strongly enhanced compared to the WT inflammasome. As a comparison, Y623A is capable of both self-oligomerizing in vitro for high ATPase activity and inducing inflammasome complex formation in cells for strong caspase-1 cleavage. Taken together, the ATP binding observed in the twisted conformation compacts NLRP6 domains and maintains NLRP6 in the inhibitory state. The structure-guided mutagenesis verified that disruption of the ATP-NLRP6 interaction facilitates higher-order assembly, stimulates the ATPase activity and promotes NLRP6 inflammasome activation in cells. NBD contributes to the inhibition of NLRP6 inflammasome Above overall structural comparison uncovered a unique mode of ATP binding in NLRP6 (Fig. 2a,b). To identify the specific structural domain that distinguishes the conformation and function of NLRP6 from other NLRs, we performed the detailed structural alignment of the NACHT domain (Extended Data Fig. 8a–d), which was reported to be an essential hub to coordinate nucleotides39–44. The structures of the NACHT domain among NLRP3, NLRP1 and NLRC4 are well aligned (Extended Data Fig. 8a). However, when we performed the alignment of the NACHT between NLRP6 twisted conformation and NLRP3, NLRP1 and NLRC4 respectively, the main discrepancy came from the NBD domain: NLRP6NBD adopts a unique orientation to interact with HD1/WHD/HD2 domains without the involvement of nucleotides at the canonical catalytic site (Extended Data Fig. 8b–e). This unique conformation generates a more compact NBD–HD1/WHD interface in NLRP6 than that seen in NLRP3, NLRC4 and NLRP1. The ADP or ATP-γS seen in other NLR structures would potentially lead to steric clashes in NLRP6 with the α3–α4 loop of the HD1 domain or the junction of NBD and HD1 domain (zoomed images, Extended Data Fig. 8b–d). Therefore, we speculated that cooperation of ‘inhibitory ATP’ (Fig. 2d) and the NBD lock the overall conformation, occluding the ‘substrate ATP’ from the canonical catalytic site (Extended Data Fig. 8b–d), thereby dampening ATPase activity and suppressing NLRP6 inflammasome activity (Fig. 2e,g). We then performed detailed interface analysis to investigate the role of the NBD in the twisted conformation. Structurally, we identified four main interaction interfaces, including NBD–WHD, NBD–HD1 and NBD–HD2 (Fig. 3a–e). The NBD–WHD interaction covers an extensive surface that can be further divided into interface 1 and interface 2 (Fig. 3b,c). In NBD–WHD interface 1, R132 and K213/K214 from NBD establish salt bridges with side chains of E492 and E484, as well as the hydrogen bonds with the backbone of K483 in the WHD domain. K483 and Q496 in WHD form hydrogen-bonding networks with the backbone of C237, the side chains of R327 and D279 in NBD (Fig. 3b). In the NBD–WHD interface 2, D499WHD potentially interacts with the backbone of A328NBD, whereas E242NBD is anchored by the backbones of R456 and Q459 in the WHD domain (Fig. 3c). In the NBD–HD1 interaction interface, R347NBD mediates hydrogen-bond formation with the backbones of C385HD1 and F386HD1. D351 from the helix 1 of HD1 forms a hydrogen-bond triad with R333 and G336 of the NBD domain. L151NBD and Y357HD1 potentially contribute to the interaction through hydrophobic contacts (Fig. 3d). NBD and HD2 encompass a shallower interface. R306NBD stabilizes the conformation of the acidic loop through interaction with backbone of E609 (Fig. 3e). The side chain of E609 does not contribute to the autoinhibition (Fig. 2g). The previously analyzed ATP-binding interface residue H539 is located at the intersection of several structural elements including HD2, WHD, NBD and the ATP molecule, further endorsing its crucial role in mediating autoinhibition (Fig. 3e). Next we conducted structure-guided mutagenesis to analyze the functional importance of these interface residues by examining the activity of cellular reconstituted NLRP6 inflammasome as described above. K213A/K214A, E242A, K483A and Q496A mutations increased the level of p20 cleavage notably (Fig. 3f), suggesting that these residues are key factors in maintaining the inhibitory conformation through NBD–WHD interaction. The reconstitution of R132A, E492A, D279A, R327A, D499A, A328G and A329G mutants did not release the inhibition based on our cellular assay (Fig. 3f). For the NBD–HD1 interface, R333, R347 and D351 are the key residues that stabilize the inhibitory conformation of NLRP6 (Fig. 3g). Mutagenesis of the NBD–HD2 interface revealed that, in addition to the mutation of critical bridge residue H539L, R306A also activates NLRP6 inflammasome in cells (Fig. 3g). In summary, the NBD domain of NLRP6 in the twisted conformation directly interacts with WHD, HD1 and HD2 to confer inhibition of NLRP6 inflammasome.Fig. 3NBD interacts directly with other subdomains of NACHT and inhibits NLRP6.a, Overview of NLRP6 structure in the twisted conformation. Principal interfaces with the involvement of NBD domain are labeled (I–IV) by black boxes and presented in detail in b–e. Domains and ATP are color coded. b–e, Close-up views of NBD–WHD interface 1 (b), NBD–WHD interface 2 (c), NBD–HD1 interface (d) and NBD–HD2 (e) interface in relation to a. Residues and ATP are shown as sticks. The domains are in cartoon representation following the color code in a. f, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT and NBD–WHD interface mutants inflammasomes in HEK-293T cells. KKAA, K213A/K214A. g, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT, NBD–HD1 interface mutants (left) and NBD–HD2 interface mutants (right) inflammasomes in HEK-293T cells. These experiments were repeated independently three times.Source data NBD interacts directly with other subdomains of NACHT and inhibits NLRP6. a, Overview of NLRP6 structure in the twisted conformation. Principal interfaces with the involvement of NBD domain are labeled (I–IV) by black boxes and presented in detail in b–e. Domains and ATP are color coded. b–e, Close-up views of NBD–WHD interface 1 (b), NBD–WHD interface 2 (c), NBD–HD1 interface (d) and NBD–HD2 (e) interface in relation to a. Residues and ATP are shown as sticks. The domains are in cartoon representation following the color code in a. f, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT and NBD–WHD interface mutants inflammasomes in HEK-293T cells. KKAA, K213A/K214A. g, Immunoblots of caspase-1 processing in the reconstituted NLRP6 WT, NBD–HD1 interface mutants (left) and NBD–HD2 interface mutants (right) inflammasomes in HEK-293T cells. These experiments were repeated independently three times. Source data Repositioning of the acidic loop modulates the NLRP6 inflammasome The structural and biochemical analysis of ATP and the NBD domain highlighted the twisted conformation as an inactive state. To elucidate the role of the extended conformation (Fig. 1d), we overlaid the two monomeric structures together and observed notable structural differences in HD1, HD2 and LRR regions (Extended Data Fig. 9a,b). Among them, the acidic loop exhibits the most significant structural rearrangement (Extended Data Fig. 9c), indicating it may have an important regulatory role between two states. In the twisted conformation, the short acidic loop (E609–N616) at the C-terminal of HD2 domain acts as a shield to stabilize the ATP molecule in the inhibitory pocket (Fig. 4a). In addition, the loop is also positioned perfectly at the cavity formed between HD2 and LRR through a multi-hydrogen-bonding network, including the triad formed among E613, H539 from HD2 and the γ-phosphate of the ATP molecule and the tetrad generated by E611, N616, adenosine ring of ATP and the backbone of R648 from LRR (Fig. 4a). The reconstituted E611A and E613A mutants with other inflammasome components yielded much higher level of caspase-1 cleavage (Fig. 2g and Fig. 4b). Furthermore, the western blot results of N616A further supported the critical roles of the acidic loop in the inactive conformation by gluing ATP, HD2 and LRR together, whereas E614 and Q535 did not contribute to the inhibition (Fig. 4b). In the extended conformation, ATP is absent from the structure. The acidic loop protrudes toward a surface that is formed by the loops connecting alpha-helix and beta-strand repeats in the LRR domain (Fig. 4c). This loop is much longer (T594–N616) than that observed in the twisted conformation, and more negatively charged glutamate residues are stabilized by the LRR domain. In fact, we observed an extensive interaction surface exhibiting charge complementarity between acidic residues on the loop and basic residues on the LRR domain (Fig. 4c). Next, we asked whether the loop–LRR interaction modulates NLRP6 activity. When we mutated several glutamate residues, including E613A, named as loop-7Mut (E601/604/607/608/609/610/613A), the loop can neither bind LRR nor stabilize the ATP-binding pocket, and western blot analysis showed strong activation of NLRP6 inflammasome (Fig. 4d). We then mutated the positively charged residues in the LRR domain that disrupt loop–LRR interaction but maintain a functional acidic loop. The results showed that R648A/R650A and R678A mutants substantially impaired NLRP6 inflammasome activity, whereas T594A/E595A, K739A and R767A mutation had minimal effect (Fig. 4d). The above structural analysis and biochemical validation suggest a model in which the acidic loop harbors the conformational plasticity between two states to modulate NLRP6 activity: the loss of loop–ATP–HD2–LRR interaction in the twisted conformation activates NLRP6, whereas the disruption of loop–LRR in the extended conformation probably pushes the loop back to the ATP-binding position, thus inhibiting NLRP6. The full loop mutant loop-7mut mimics a possible third conformation with the fully flexible loop, which may lead to full activation of NLRP6 through higher-order assembly. In other words, we speculate that the extended conformation may be an intermediate state between the auto-inhibited twisted conformation and the fully activated higher-order oligomers.Fig. 4Conformations of the acidic loop modulate NLRP6 inflammasome activity.a, Left: overview of the NLRP6 twisted conformation. Right: zoomed-in image of loop-mediated ATP binding and domain stabilization. Interaction residues are highlighted in stick representation. ATP is highlighted in ball-and-stick representation. b, Immunoblots of caspase-1 processing to elucidate loop-mediated interactions in a. c, Left: overview of NLRP6 extended conformation. Right: zoomed-in image of loop–LRR interaction. Interaction residues are highlighted in stick representation. d, Immunoblots of caspase-1 processing to elucidate the loop–LRR interaction in c. e, Structure-derived model of NLRP6 inflammasome inhibition and activation. Cartoon blocks, different domains; hexagon, ATP; curved lines, acidic loop from both conformations. The cartoon locks on top suggest different activating states: inhibited (left, locked), intermediate (middle, partially open) and activated (right, fully open) state. Experiments in b and d were repeated independently three times.Source data Conformations of the acidic loop modulate NLRP6 inflammasome activity. a, Left: overview of the NLRP6 twisted conformation. Right: zoomed-in image of loop-mediated ATP binding and domain stabilization. Interaction residues are highlighted in stick representation. ATP is highlighted in ball-and-stick representation. b, Immunoblots of caspase-1 processing to elucidate loop-mediated interactions in a. c, Left: overview of NLRP6 extended conformation. Right: zoomed-in image of loop–LRR interaction. Interaction residues are highlighted in stick representation. d, Immunoblots of caspase-1 processing to elucidate the loop–LRR interaction in c. e, Structure-derived model of NLRP6 inflammasome inhibition and activation. Cartoon blocks, different domains; hexagon, ATP; curved lines, acidic loop from both conformations. The cartoon locks on top suggest different activating states: inhibited (left, locked), intermediate (middle, partially open) and activated (right, fully open) state. Experiments in b and d were repeated independently three times. Source data In addition to the conformational change of the acidic loop, the movements of other structural elements were also analyzed (Extended Data Fig. 9b,d–f). A383 from the HD1 domain, V649 from the LRR domain and V541 from the HD2 domain moved toward the ATP molecule during the transition from extended to twisted conformations (Extended Data Fig. 9d–f), collectively making the ATP-binding domains closely connected in the inactive state. Further cell-biology validation confirmed that when we introduced A383G, A383P and V541G mutations to break local secondary structure, NLRP6 became constantly active. However, the V649G mutation did not change NLRP6 activity (Extended Data Fig. 9g). Discussion In summary, the cryo-EM structures of NLRP6 monomer reveal the structural basis of NLRP6 inhibition by ATP and provide important insights into how NLRP6 activity is regulated by specific structural elements under different states. First, in the twisted conformation, ATP is a critical ‘molecular lock’ to bring HD1, WHD, HD2, LRR and the acidic loop together, and further stabilize the conformation of NBD domain, thus restricting NLRP6 into a compact and inactive state. The spontaneous activation of NLRP6 was observed by introducing ATP-binding interface mutants, as indicated by the ATP hydrolysis assay and reconstituted NLRP6 inflammasome in HEK-293T cells. ATP–ADP exchange was reported to be involved in the conformation switch of NLR proteins between inactive and active states39,40. Unlike other solved NLR structures that have ADP or ATP-γS occupying the classical enzyme catalytic pocket, the twisted conformation of NLRP6 uses a noncanonical pocket for ATP binding and mediates subsequent inhibition, presenting an unprecedented mechanism of NLR protein regulation. The structural and biochemical analysis implicates an elegant ‘ATP-regulated ATP hydrolysis and NLRP6 inflammasome activation’ model. Under steady state, high cellular ATP concentration47 may prevent NLRP6 autoactivation through ATP binding to the newly identified inhibitory pocket. Conversely, pathogen challenges or damage signals may impact the mitochondria functions and decrease the ATP level sharply in cells. Depletion of ATP may activate NLRP6 inflammasome, and further promote ATP hydrolysis at the catalytic site. The continuous consumption of ATP generates a positive feedback loop for NLRP6 activation. In this model, ATP acts as a molecular rheostat to tune the NLRP6 inflammasome activity through dual regulation at ATP inhibitory and ATP catalytic sites. An analogous mechanism has been observed in the case of sterile alpha and Toll/interleukin-1 receptor (TIR) motif-containing protein 1 (SARM1)—a central regulator of the injury-induced axon degeneration48,49. In the resting state, nicotinamide adenine dinucleotide (NAD+) is the inhibitor of SARM1 by binding to the armadillo repeat (ARM) domain50 (inhibitory site). Sterile injury activates SARM1 and switches its conformation. Activated SARM1 rapidly depletes NAD+ through TIR domain oligomers51,52 (catalytic NADase site), thus generating a feedback loop for SARM1 activation. The structural alignment of NLRP3 in complex with MCC950 (refs. 41,53) and NLRP6 in complex with ATP showed these two small molecules share similar binding locations (Extended Data Fig. 10a). Sequence alignment also showed that several small-molecule binding residues are highly conserved between these two proteins (Extended Data Fig. 10b). Although MCC950 is a potent NLRP3 inhibitor, NLRP6 inflammasome activity was not inhibited by MCC950 in our reconstituted system (Extended Data Fig. 10c). In contrast, there exhibited slight increase of p20 level at high concentration (50 μM and 250 μM) of MCC950 (Extended Data Fig. 10c). The small-molecule selectivity of the binding pockets opens an exciting pharmacological avenue to discover NLRP6-specific inhibitors/activators in the future direction. Second, in the twisted state, the NBD domain of NLRP6 adopts a unique conformation different from the NBDs of inactive NLRP3, NLRP1 and NLRC4. The interactions between NBD and WHD/HD1/HD2 are direct and mediated predominantly by electrostatic contacts. Although we do not know the advantage of this distinct conformation in NLRP6NBD, we can speculate that this conformation facilitates inhibition through two aspects: (1) the tight interface between NBD and WHD/HD1 prevents ‘substrate ATP’ binding to the potential canonical catalytic site due to steric hindrance, thereby preventing ATP hydrolysis and subsequent activation (Fig. 3b–d and Extended Data Fig. 8b–d); (2) the NBD conformation cooperates with the ‘inhibitory ATP’ and HD2 to stabilize the inactive state (Fig. 3e). We cannot exclude that NLRP6 may adopt an ATP hydrolysis conformation given that ATP hydrolysis activity is detected and regulated in our study. Mutagenesis on Walker A/Walker B motifs (classical ATPase catalytic motifs) did not abolish enzyme activity compared to WT (Fig. 2c), raising the possibility of having a completely new site or a new conformation for ATPase in NLRP6. This possibility is also implicated by the sequence alignment54 of the important ‘ADP-sensor’ residue— a highly conserved histidine (H443 in NLRC4/H522 in NLRP3/H623 in NLRP1) at the WHD domain of other NLR inflammasome receptors. This histidine plays critical roles in coordinating ADP to stabilize the inactive conformation and may regulate ADP–ATP exchange at the canonical catalytic site. However, NLRP6 has a nonconserved aspartic acid (D499) at the corresponding location54, which is likely to disrupt ADP binding. This uniqueness might partially explain why NLRP6 does not share the stringent canonical ATPase pocket observed in other NLRs. Notably, the higher-order form of NLRP6 exhibited elevated ATPase activity (Extended Data Fig. 6a). Whether the ATP hydrolysis conformation is stabilized by the oligomeric assembly of NLRP6 awaits further structural investigation. Third, the acidic loop at the connecting region of HD2 and LRR domain is an exclusive feature of human NLRP6 and mouse NLRP6. Our structure-based cellular studies suggested that the conformational switch of the acidic loop between different states plays a crucial role in modulating NLRP6 inflammasome activity. Overall, our structure determination and biochemical exploration reveal a multi-element involved hypothetical model of NLRP6 inflammasome inhibition and activation (Fig. 4e): the twisted conformation is the inhibitory state stabilized by the cooperative regulation of ATP, the acidic loop and NBD (Fig. 4e, left); the extended conformation, lacking ATP and with an altered NBD that is not detectable by cryo-EM represents an intermediate state that may have mutual transition with the inhibited state (Fig. 4e, middle). The model is supported by our structural observation of two monomeric conformations from the same gel-filtration fraction. The higher-order assembly observed in our purification profile (Extended Data Fig. 1a,c) has high levels of ATPase activity (Extended Data Fig. 6a–d), suggesting the existence of a fully activated state (Fig. 4e, right). Our mutagenesis study showed that disruption of ATP-/NBD-binding favored the activated state. However, the physiological trigger for the conversion of the extended conformation to the fully activated state is unknown. A possible explanation is that the extended conformation is prone to ligand-mediated or protein concentration-dependent activation. From the insight of our previous discovery21, the fully flexible NBD not seen in the extended conformation may be an ideal module to mediate multivalent interaction with NLRP6 ligands (dsRNA/lipoteichoic acid) and promote the formation of biomolecular condensates, thus increasing the local concentration and facilitating the higher-order assembly of NLRP6. We acknowledge that our overexpression system in reconstituted HEK-293T cells has its caveats and limitations, such as potential abnormal expression level (significantly higher amount of inflammasome components than the endogenous level) and expression in a cell type that does not usually express NLRP6 (expressed predominantly in the gastrointestinal tract). In future study, to further explore the physiological relevance of our findings, we will use physiological cell lines such as primary intestinal epithelial cells (with endogenous expression of NLRP6 inflammasome components) to study the functional significance of ATP/NBD-binding disruption, primarily through ATP-binding pocket/NBD-interface mutants generated from CRISPR–Cas9 knock in, and validate NLRP6 inflammasome activation. In conclusion, as a dominant inflammasome sensor in the gastrointestinal system, NLRP6 plays pivotal roles in regulating intestinal inflammation and its overactivation is related tightly to inflammatory bowel disease. Our structural discovery of human NLRP6 inhibition by ATP provides important insights into the design of therapeutics to target chronic inflammatory diseases. Methods Plasmid construction Human NLRP6 with an N-terminal flexible region deleted (amino acids 14–892) was cloned into pFastBac-HTB vector engineered with an N-terminal 6×His-MBP tag. Full-length human NLRP6, human ASC and human caspase-1 were cloned into the pcDNA3.1 vector with an HA-Flag tag at the N terminus and a myc tag at the C terminus. All the NLRP6 mutants in the biochemical and cellular studies were generated through site-direct mutagenesis with the QuikChange strategy. The original human NLRP6 full-length and human caspase-1 plasmids were reported in our previous study21,28. The human ASC plasmid was purchased from Addgene (cat. no. 73952). All plasmids were verified by DNA sequencing. Protein expression and purification NLRP6 (14–892) was expressed in SF9 cells (Thermo Fisher, cat. no. 11496015). The baculovirus of NLRP6 was generated using the Bac-to-Bac system (Invitrogen). For expression, an 800-ml suspension of cultured SF9 cells (density, 2.0 × 106 cells ml−1 in HyClone SFX-Insect cell medium) was infected with 16 ml baculovirus (multiplicity of infection of 3) at 27 °C. After culturing at 27 °C for 48 h, cells were collected, resuspended in a buffer containing 20 mM HEPES pH 7.4, 300 mM NaCl, 20 mM imidazole pH 7.4 (Buffer A) and lysed by sonication. After centrifugation at 186,000g for 45 min, the supernatant was incubated with 1 ml Ni-NTA resin (Qiagen) for 45 min. After incubation, the resin was loaded onto a gravity column and washed with 25 column volumes of Buffer A. The bound MBP-tagged NLRP6 was eluted with 20 mM HEPES pH 7.4, 300 mM NaCl, 300 mM imidazole pH 7.4 (Buffer B) with a volume of 3 ml. The eluted product was pooled, concentrated to 0.5 ml (Amicon Ultra centrifugal filters, 4 ml, molecular weight cutoff 30 kDa) and further purified by SEC (Superdex 200 increase 10/300 GL column, Cytiva) pre-equilibrated with a buffer of 20 mM HEPES, pH 7.4, 300 mM NaCl and 1 mM dithiothreitol (DTT). The eluates were collected at intervals of 0.5 ml. Three fractions corresponding to the 12.2-ml peak were pooled and subjected to a second round of gel filtration in a lower salt buffer (20 mM HEPES, pH 7.4, 200 mM NaCl and 1 mM DTT). The peak fractions were concentrated for further cryo-EM data acquisition and biochemical assays. All the purification steps were conducted at 4 °C. Negative-staining EM For negative staining, 5 μl of NLRP6 void fraction (concentration: 0.05 mg ml−1) or NLRP6 monomeric fraction (concentration: 0.015 mg ml−1) was applied to a glow discharged copper grid (Ted Pella, cat. no. 01754F), incubated for 1 min and absorbed with filter paper. The grid was stained with 5 μl 2% uranyl acetate solution (Electron Microscopy Sciences, cat. no 22400-2) for two rounds (first round 30 s; second round 1 min) and air dried. Images were acquired using a JEOL JEM-1400 120-kV transmission electron microscope equipped with an AMT XR111 high-speed 4,000 × 2,000 pixel phosphor-scintillated 12-bit CCD camera (Washington University Center for Cellular Imaging). ATPase activity assay The ATPase activity assay was carried out using an ATPase/GTPase activity assay kit (Sigma, cat. no. MAK113) following the manufacturer’s instructions. For ATPase assays of monomeric and void samples of NLRP6 WT and mutants (ATP-binding interface mutants, Walker A/B motif mutants), 10 µl protein was added into a 96-well plate containing 10 µl ATP and 20 µl of assay buffer (40 mM Tris, 80 mM NaCl, 8 mM MgAc2, 1 mM EDTA, pH 7.5) in each well and mixed to initiate the reaction. This reaction system’s final protein and ATP concentrations are 2 µM and 1 mM, respectively. After 90 min of incubation at room temperature, 200 µl of malachite green reagent was added to each well to stop the enzyme process for 30 min and produce the colorimetric result. For the NLRP6 concentration-dependent ATPase assay, NLRP6 monomer and BSA concentrations of 0, 0.47, 0.94, 1.88, 3.75 and 7.5 µM and the void concentrations of 0, 0.24, 0.47, 0.94, 1.88 and 3.75 µM were used. The reaction time was 90 min. For the time-dependent ATPase assay, timepoints of 0, 30, 60, 90 and 120 min were used. The protein concentration was set to 2 µM. The optical density of all wells was measured using a spectrophotometric multi-well plate reader set to 620 nm. Cryo-EM sample preparation and data collection The human NLRP6 protein was deposited onto a pre-glow-discharged copper grid with holey carbon films (QUANTIFOIL, 300-mesh, R1.2/1.3, cat. no. N1-C14nCu30-01), blotted for 4 s under 100% relative humidity with an applied force of 1, and then plunged rapidly into liquid ethane using a Vitrobot (Thermo Fisher Scientific). Cryo-EM data for human NLRP6 were collected at 300 kV on a TFS Titan Krios microscope equipped with a Falcon4EC direct detector (NCCAT/NYSBC). The movies were collected in electron event representation mode with a total dose of 48.44 e− Å−2 with an exposure time of 6 s. Automated single-particle data acquisition was performed using Leginon56 at a nominal magnification of ×165,000 and with a calibrated pixel size of 0.73 Å. The images were recorded at a defocus range of 0.6–2.5 µm. Cryo-EM data processing A total of 14,261 micrographs of human NLRP6 were collected and processed using CryoSPARC v4.4.0 (ref. 33). Initial motion correction was performed with MotionCor2 (ref. 57), followed by Patch contrast transfer function estimation to determine contrast transfer function values for each micrograph. A total of 7,565,037 particles of human NLRP6 were picked and extracted with a box size of 300 pixels. All particles underwent reference-free 2D alignment in CryoSPARC, yielding 1,005,509 particles that were selected for ab initio reconstruction and heterogeneous refinement. No symmetry was applied during any of the classification steps. Subsequently, 682,063 particles of human NLRP6 extended conformation and 257,383 particles of human NLRP6 twisted conformation were selected and subjected to several rounds of 2D classification, heterogeneous refinement and nonuniform refinement in CryoSPARC. Approximately 6.57% of particles were classified as background noise at the three-dimensional steps. The data processing yielded the final cryo-EM maps for NLRP6 twisted conformation at 3.7 Å and NLRP6 extended conformation at 3.8 Å. The resolution was determined using the gold-standard Fourier shell correlation criterion (threshold = 0.143) and evaluated with ResMap58. A summary of data processing is provided in Table 1 and Extended Data Fig. 2. Model fitting and refinement For the model building of NLRP6 extended conformation, the structural domains (HD1–WHD–HD2–LRR: 349–892) from the AlphaFold2 (ref. 34) model (AF-P59044-F1) of human NLRP6 were fitted into the cryo-EM densities using ChimeraX59. For the model building of NLRP6 twisted conformation, besides the docking from HD1 to LRR domains, the NBD domain (118–348) was built individually. The acidic loops in both conformations were built manually based on the density. Manual adjustments were performed subsequently in Coot60, followed by real-space refinement in Phenix61. The cryo-EM maps and atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB), respectively. Data collection and refinement statistics are provided in Table 1. All structural representations were created with ChimeraX59 and PyMOL62. Nucleotide extraction test using LC–MS analysis Protein samples were prepared in Buffer S (20 mM HEPES, 300 mM NaCl, 1 mM DTT, 10 mM MgCl2). Purified recombinant NLRP6 at 0.5 mg ml−1 to yield 92.5 μg total protein was used for the test, along with 100 μl of 0.2 mg ml−1 (20 μg) of ATP standard tested separately in parallel. An equal volume of 10% TCA was added to achieve a final concentration of 5% TCA (pH <2), and samples were incubated on ice for 10 min. Because low-concentration protein was precipitated incompletely, the TCA treated sample was processed using a 50-kDa molecular weight cutoff centrifugal filter to remove proteins, followed by an additional wash with 100 μl of 1% TCA. Off-line MAX–SPE was performed to enrich and desalt adenosine nucleotides. Samples were subjected to four sequential methyl tert-butyl ether:water liquid–liquid extractions to remove excess TCA, neutralized with a drop of 10% ammonium hydroxide, and loaded onto conditioned MAX–SPE cartridges. Cartridges were washed twice with 1 ml of 1% ammonium hydroxide in acetonitrile and eluted twice with 2% formic acid in methanol. Eluates were dried without heating and reconstituted in 30 μl of 80% acetonitrile containing 10 mM ammonium acetate. LC was performed using a Vanquish Horizon UHPLC system (Thermo Scientific) equipped with a Waters Acquity BEH Amide column (2.1 × 150 mm, 1.7 μm) using a 9 min gradient. Mobile phase A consisted of 20 mM ammonium acetate in water (pH 9.0 adjusted with ammonium hydroxide) and mobile phase B was acetonitrile (pH 9.0 adjusted with ammonium hydroxide). Samples were transferred to autosampler vials and 5 μl was injected. MS was performed on an Exploris 240 Orbitrap mass spectrometer (Thermo Scientific) operated in negative ion mode. Adenosine nucleotides were analyzed by full-scan MS1 acquisition followed by parallel reaction monitoring acquisition for MS2. Data processing was performed using Skyline (version 25.1). High-resolution extracted ion chromatograms were generated for ATP (m/z 505.9885) within 5 ppm error, and abundances were quantified by integrating precursor peak areas. All sample preparation steps were verified using ATP standards processed in parallel. Cell culture HEK-293T cells (ATCC, cat. no. CRL-11268) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with L-glutamine and 10% fetal bovine serum (FBS) at 37 °C in a humidified 5% CO2 incubator. NLRP6 inflammasome assay To reconstitute the human NLRP6 inflammasome, HEK-293T cells were seeded in 12-well tissue culture plates at 1 × 105 cells per well. After 36 h of incubation (reaching ~70% confluency), cells were cotransfected with the following plasmids using Lipofectamine 2000 (Invitrogen) and Opti-MEM reduced serum medium: FLAG-tagged NLRP6 (wild type, R132A, L151A, K213A/214 A, C237L, E242A, D279A, R306A, R327A, A328G, A329G, A330G, R333A, R347A, D351A, Y357A, E379A, F382A, A383G, A383P, F386A, K483A, L485P, E492A, Q496A, D499A, E504A, Q535A, H539L, V541G, R545A, T594A/E595A, E601/604/607/608/609/610/613 A, E609A, E611A, E613A, E614A, N616A, E620A, Y623A, R648/650 A, V649G, D655A, R678A, K739A or R767A; 300 ng), FLAG-tagged ASC (25 ng) and FLAG-tagged caspase-1 (300 ng). Following 24 h of transfection, cells were collected, washed three times with phosphate-buffered saline (PBS), and lysed for subsequent immunoblot analysis of FLAG-tagged proteins and caspase-1 activation (p20 band). For MCC950 treatment, 4 h after transfection, cells were treated with MCC950 (stock dissolved in dimethylsulfoxide) for 20 h and harvested for immunoblotting, with an equivalent volume of dimethylsulfoxide used as control. Immunoblotting Cell pellets were lysed in 120 μl of 1× SDS sample buffer and denatured by boiling at 95 °C for 10 min. Proteins were separated on Bolt 4–12% Bis-Tris Plus WedgeWell gels (1.0 mm; Invitrogen) at 160 V for 40 min and then transferred to nitrocellulose membranes using the Power Blotter Select Transfer Stacks and Station System (Invitrogen). Membranes were blocked with 5% nonfat milk in PBS for 1 h at room temperature, followed by incubation with primary antibodies (1:1,000 dilution in blocking buffer) at 4 °C for 36 h. After three washes with PBS containing 0.1% Tween-20, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:2,000 dilution) for 1 h at room temperature. Following three additional washes with PBS containing 0.1% Tween-20, membranes were rinsed and visualized using an iBright CL1500 Imaging System (Invitrogen). Primary antibodies used were anti-Flag (Sigma, cat. no. F1804; Proteintech, cat. no. 20543-1-AP), anti-caspase-1 (Cell Signaling Technology, cat. no. 3866S), anti-β-actin (Cell Signaling Technology, cat. no. 4970S). Secondary antibodies used were anti-mouse-HRP (Invitrogen, cat. no. 31430) and anti-rabbit-HRP (Cell Signaling Technology, cat. no. 7074S). Statistics and reproducibility No statistical method was used to predetermine sample size. All quantitative data were presented as the mean ± s.e.m. Exact P values are provided in Supplementary Tables 1 and 2. No data were excluded from the analyses. Randomization and blinding methods were not used. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-026-01878-5. Supplementary information Supplementary InformationSupplementary Tables 1 and 2. Reporting Summary Peer Review File Supplementary Tables 1 and 2. Reporting Summary Peer Review File Source data Source Data Fig. 2Unprocessed western blots and/or gels. Source Data Fig. 3Unprocessed western blots and/or gels. Source Data Fig. 4Unprocessed western blots and/or gels. Source Data Fig. 2Statistical source data. Source Data Extended Data Fig. 1Unprocessed western blots and/or gels Source Data Extended Data Fig. 1Statistical source data. Source Data Extended Data Fig. 6Statistical source data. Source Data Extended Data Fig. 7Unprocessed western blots and/or gels. Source Data Extended Data Fig. 7Statistical source data. Source Data Extended Data Fig. 9Unprocessed western blots and/or gels. Source Data Extended Data Fig. 10Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Statistical source data. Unprocessed western blots and/or gels Statistical source data. Statistical source data. Unprocessed western blots and/or gels. Statistical source data. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels.